Why AI isn’t booking your flights – yet

FOR an industry that prides itself on innovation, commercial aviation’s digital interfaces have changed remarkably little.

Whether you’re booking a S$15,000 first-class fare from Singapore to London or a S$70 low-cost hop to Bali, the process looks strikingly similar to what it did in 1998. The Internet Booking Engine remains frozen in time, merely refreshed by better visuals and faster search results. Step one: enter departure and arrival airports. Step two: pick your dates. Step 3: select passengers, then scroll through flight options, fares, ancillaries, passenger details, and finally payment.

Three decades on, we’ve seen stunning leaps in digital capability: self-driving taxis on public roads, artificial intelligence (AI) co-pilots in corporate offices, and personalised retail ecosystems that can anticipate your next purchase. Yet booking a flight remains a multi-step ritual that’s stubbornly procedural, transactional and unintelligent.

Why? And more importantly, why do travellers still hesitate to let AI do what it already can?

The great expectation: conversational travel

The vision of AI-driven booking has been around since at least the early 2010s. The dream was elegantly simple: natural-language travel planning.

You’d tell an intelligent assistant: “Book me flights from Frankfurt to New York, from Oct 10 to 20; I prefer Star Alliance carriers; budget up to 1,000 euros; direct is best, but connections under three hours are fine.”

The assistant would parse your intent, search across multiple inventories, weigh fares, schedules and preferences, and then – crucially – decide for you. No drop-down menus, no endless scrolling. Just one seamless conversational flow between human intent and algorithmic execution.

In theory, it’s entirely achievable today. In practice, almost no one books this way.

Two barriers: capability and trust

The first reason is straightforward: There is simply no widely deployed, functioning solution.

Airlines, online travel agencies (OTAs), and travel management companies have not embedded natural-language AI into their booking engines at scale. Despite major advances in large language models, few players have built systems that can translate conversation into structured, compliant, International Air Transport Association-standard bookings.

Beyond inertia, it’s also economics. Booking engines operate on razor-thin margins and are deeply integrated with legacy systems. Rebuilding that architecture around generative AI is not a user-interface problem – it’s a business-model problem.

But even if the technology existed, the second barrier would remain: trust.

Travellers do not trust the results. They fear losing control of the decision-making process.

When AI suggests a flight, most users still check again, against other platforms. Experience has taught them inconsistency: identical searches on different devices, at the same time, often produce different prices and itineraries. In a category where price remains the strongest decision factor, the fear of “not getting the best deal” outweighs the convenience of delegation.

The psychology of control

Booking travel is also deeply psychological. Travellers want agency – the feeling that they chose their fare, seat, layover. They compare options not just for price, but also for validation that they got the best value.

AI threatens to remove that. Even if it delivers an objectively optimal itinerary, most people will still question it – because they didn’t see the trade-offs themselves.

The irony then is that AI’s advantage – speed and efficiency – becomes its liability. An assistant that instantly “finds the best option” denies the user the cognitive process of exploration. And in consumer psychology, exploration is part of satisfaction.

A study by BAA & Partners on personalisation found that travellers are 65 per cent more likely to complete a booking when they perceive transparency in choice, even if the final fare is higher. The booking interface is also a reassurance mechanism.

Inspiration vs transaction

Before trust even enters the equation, there’s another challenge: inspiration.

Many trips begin with an idea. A business trip that could be extended. A family reunion somewhere warm. AI today is not yet good at contextual inspiration – the ability to understand that “a beach holiday with a four-hour flight limit and reliable weather” might mean different things to different people.

Search engines and OTAs are still optimised for transaction, not discovery. They expect you to know your destination. That’s a fundamental limitation when building conversational AI – because language itself is exploratory. The phrase “find me somewhere relaxing, not too expensive, ideally with nature” is linguistically simple, but computationally complex.

Without rich behavioural data and contextual memory, such as what you liked the last time and what you can afford, AI assistants cannot yet mimic a human travel agent’s intuition. Until that gap closes, AI will remain better at automating decisions than inspiring them.

The supply-side problem

There’s also a supply-side challenge. Airlines and OTAs operate in a highly fragmented, competitive ecosystem where each protects its own data and pricing logic. Generative AI thrives on data integration, but in travel, integration is almost a taboo.

Airlines fear commoditisation if AI platforms aggregate and normalise their offers too transparently. OTAs fear disintermediation. Global distribution system providers guard their application programming interfaces as profit centres. The result: an opaque marketplace where price discrepancies, caching delays, and fare rules vary by milliseconds.

When users see different fares across channels, trust erodes in the system itself. AI cannot fix a trust problem built into the industry’s plumbing.

A glimpse of what’s next

And yet, change is coming. Slowly.

Some early adopters – notably in hospitality – are testing generative AI as a “concierge layer” rather than a booking engine. Expedia’s ChatGPT plug-in, for example, helps users plan but still redirects them to a traditional interface for payment. Airlines such as Lufthansa and Emirates are experimenting internally with AI-driven customer assistance, but not yet with full conversational bookings.

The path forward may not be replacing the booking interface, but augmenting it. Think of a hybrid model: AI acts as an intelligent adviser, narrowing down thousands of permutations into a few high-quality, explainable choices. The traveller still sees options – but with contextual reasoning (“This fare is cheaper because it departs earlier and avoids a long layover”).

That transparency could rebuild trust. It also mirrors how a good human travel agent operates – explaining trade-offs, not hiding them.

From transactions to relationships

If airlines and OTAs want travellers to trust AI, they must think more about the relationship. Trust in AI doesn’t emerge from a one-time perfect search; it’s earned over repeated, consistent outcomes. The user must feel that the system “understands” them.

This requires a fundamental redesign of digital booking systems: from static, session-based transactions to dynamic, data-driven relationships. Imagine an airline platform that recalls your preferences, connects them to real-time pricing, and proactively alerts you when your ideal itinerary fits your budget – before you even ask.

That’s not science fiction. It’s what advanced personalisation engines in e-commerce already do. Travel, ironically, lags retail by nearly a decade in personalisation maturity.

Final descent: trust takes time

AI booking will succeed not when it works, but when it’s trusted. And that trust will come from four converging shifts:

Transparency of logic: Users must understand why a fare was chosen.

Consistency of data: Fare parity and synchronisation across channels must improve.

Personalisation memory: AI must learn from each interaction, evolving from a generic assistant to a loyal travel companion.

Human fallback: True confidence in automation often comes when users know a human can step in if needed. Hybrid experiences will build the bridge.

AI will eventually book our flights. But adoption will be evolutionary, not revolutionary. The barriers aren’t just technical; they’re emotional, behavioural and systemic.

For now, we may still prefer to do it manually. Yet the day will come when we’ll simply say “Book it”, and mean it – not because the machine got smarter, but because we learnt to trust it.

The writer is founder of BAA & Partners